EP4070426A1

EP4070426A1 - Method for operating a data centre in an electrical network, and data centre for carrying out such a method

Info

Publication number: EP4070426A1
Application number: EP20820126.9A
Authority: EP
Inventors: Thomas Kottke; Niko Mittelmeier
Original assignee: Rolls Royce Solutions GmbH
Current assignee: Rolls Royce Solutions GmbH
Priority date: 2019-12-06
Filing date: 2020-12-04
Publication date: 2022-10-12
Also published as: DE102019219111B4; US20220302702A1; CN114731042A; DE102019219111A1; WO2021110905A1

Abstract

The invention relates to a method for operating a data centre (1) in an electrical network (3), wherein - a plurality of calculation tasks (11) are processed by the data centre (1), wherein - each calculation task (11) of the plurality of calculation tasks (11) is assigned a prioritisation value and/or a calculation complexity, wherein - the calculation tasks (11) are processed taking into account their individual prioritisation value and/or calculation complexity and taking into account a power provision parameter (15) of the electrical network (3).

Description

Method for operating a data center on an electrical network and data center for carrying out such a method

The invention relates to a method for operating a data center on an electrical network, as well as a data center which is set up to carry out such a method.

Such a data center can be operated as a consumer on an electrical network, which can be at least partially fed by volatile power generators as sources of electrical power. Such volatile power generators can in particular be assigned to so-called regenerative energies, which can be photovoltaic or wind power plants, for example. What these volatile power generators have in common is that the electrical power generated fluctuates over time due to uncontrollable external influences. For a stable operation of the electrical network it is therefore necessary to provide control power. This leads to increased costs associated with the operation of such an electrical network. In addition, Lalle have become known in which certain consumers have been disconnected from the network with a short warning period by an operator of such a network or have been throttled with regard to the power that they were allowed to use. This can lead to unpredictable or at least difficult to predict difficulties in the operation of such a consumer, for example if calculation tasks to be processed at short notice cannot be processed by the computer center. If a plurality of power generators that can be controlled with regard to their output power is assigned to the power grid, it is possible that the efficiency of such power generators depends on the output power output. If necessary, these power generators are operated with reduced efficiency, especially when there is a fluctuating power demand by the consumers connected to the electrical generator.

The invention is based on the object of a method for operating a data center on an electrical network and a data center for implementing such a network To create a method, wherein the disadvantages mentioned are at least reduced, preferably do not occur.

The object is achieved in that the present technical teaching is provided, in particular the teaching of the independent claims and the embodiments disclosed in the dependent claims and the description.

The object is achieved in particular by creating a method for operating a data center on an electrical network, which is also referred to as a power network. A number of calculation tasks are processed by the computer center. Each calculation task of the plurality of calculation tasks is assigned at least one parameter selected from a prioritization value and a calculation effort; or at least one parameter selected from the prioritization value and the calculation effort is assigned to each calculation task of the plurality of calculation tasks. The calculation tasks are processed taking into account their respective at least one characteristic variable and a power supply parameter of the electrical network. This makes it possible to take into account a state of the electrical network and / or of the power generators assigned as sources - quasi proactively - when operating the data center, in particular to adapt the operation to the state of the electrical network or the state of the power generators assigned to it. The data center can thus advantageously be included in the stability control of the electrical network, preferably without the risk that computation tasks that are urgently to be processed cannot be processed. Alternatively or additionally, the processing tasks can advantageously be processed in such a way that the power generators assigned to the electrical network are operated with a high, preferably as high an efficiency as possible.

Both parameters, that is, prioritization value and calculation effort, are preferably taken into account.

An electrical network is understood here to mean, in particular, a network for supplying electrical power or energy, in particular the electrical network is an electrical energy supply network. The electrical network can be a AC voltage network or a DC voltage network. The electrical network can in particular be designed as an island network. It is particularly possible that only the data center is operated as the only consumer on the electrical network. The electrical network is then assigned to the data center, as it were, as a separate network. However, it is also possible for the electrical network to be designed as an island network, with other loads being connected to the electrical network in addition to the data center. Alternatively, however, it is also possible for the electrical network to be designed as an interconnected network or a supra-regional network.

A calculation task is to be understood in particular as an order for the calculation or processing of data by the computer center. Such a calculation task is also referred to as a task. Such a calculation task can also be a partial calculation task of a superordinate overall calculation task.

A prioritization value is understood in particular to be an identifier assigned to the calculation task, which identifies an urgency of the calculation task. In particular, the prioritization value can be a completion date on which the result of the calculation task is available or the processing of the calculation task must be completed. In particular, such a prioritization value can be specified by a customer or client of the data center.

A calculation effort is understood to mean, in particular, a time and / or energetic effort associated with processing the calculation task and / or an - possibly virtual, i.e. computational - cost effort associated with the processing of the calculation task. In particular, the computation effort can be a number of computation steps required to process the computation task, or the computation effort can depend on the number of computation steps required to process the computation task.

The prioritization value and / or the calculation effort can be specified by a client or customer of the data center, or they can be ascertained or determined before the start of the method proposed here. In this case, a processing task is a prioritization value and / or a calculation effort at the point in time assigned to the implementation of the procedure. However, it is also possible that a prioritization value and / or calculation effort - in particular from the computer center - is determined or specified for a processing task at the time the method is carried out. In this case, the prioritization value and / or calculation effort is assigned to the processing task.

A power supply parameter is understood in particular to be a parameter that characterizes a — current or future — state of the electrical network and / or of a power generator assigned to the electrical network.

That the processing tasks take into account their respective prioritization value and / or processing effort, and taking into account the

Service provision parameters are processed, means in particular that a chronological order in which the processing tasks are processed, and preferably a parallelization of the processing tasks, depending on the respective at least one parameter, i.e. prioritization value and / or calculation effort, and depending on the service provision parameter, is determined. In particular, it is preferred to specify which calculation task is processed by the computer center and at what point in time.

According to a development of the invention it is provided that the

Power provision parameter is a network stability parameter. This advantageously enables the data center to be integrated into the stability control of the electrical network. In particular, in addition to calculation tasks with a higher prioritization value and / or lower calculation effort, calculation tasks with a lower prioritization value and / or higher calculation effort are also calculated if the network stability parameter indicates that more power must be drawn from the network to stabilize the electrical network; that is, in the electrical network - possibly also in the future, in the sense of a prediction - more power is offered than is consumed, so that the consumed power is to be increased for stabilization. In contrast, only those calculation tasks with a higher prioritization value and / or lower calculation effort are preferably processed if the network stability parameter indicates that less power may be used to stabilize the electrical network; that is, in the electrical network - possibly also in the future, in the sense of a prediction - less service offered than accepted, so that the accepted service should be reduced for stabilization. This is possible, in particular, since the calculation tasks with a lower prioritization value and / or a higher calculation effort are advantageously calculated when more power can be drawn from the electrical network. The data center can thus advantageously be used as a control power sink for stabilizing the electrical network, in that it takes increased power in times of power peaks in the electrical network. However, it can also be used as a - at least virtual - control power source in that it quasi releases power requirements in times of reduced power in the electrical network, that is to say it decreases or draws less electrical power.

A higher prioritization value is understood here to mean a prioritization value which - in comparison to a prioritization value of another calculation task - indicates a higher urgency of a calculation task. In contrast, a lower prioritization value is understood to mean a prioritization value which - compared to a prioritization value of another calculation task - indicates a lower urgency of a calculation task. Similarly, the terms “higher” and “lower” are to be understood as relative information related to the calculation tasks with a view to the calculation effort.

According to a further development of the invention, the power supply parameter - in particular as an alternative or in addition to a network stability parameter - is an efficiency of a power generator for the electricity network. Particularly preferably, a plurality of efficiencies of a plurality of power generators for the power grid are considered as the power supply parameters, the calculation tasks being processed in such a way that the efficiencies of the largest possible number of power generators are as high as possible over the longest possible period. In particular, an overall efficiency of power generators connected to the electrical network is preferably optimized, in particular maximized. If it is known in which power range which power generator has its optimal efficiency, the calculation tasks can be processed accordingly, in particular with regard to their processing time, so that the power generators work in their optimal efficiency range. In particular, the processing of the calculation tasks is preferred in terms of time planned that the power generators are always operated at their optimal efficiency. This includes that at certain times only a subset of the power generators is operated, while a complementary subset of the power generators can be switched off. The number of operated power generators and the processing of the calculation tasks are preferably coordinated with one another in such a way that those power generators that are operated can always be operated at their optimal efficiency. For this purpose, in particular, calculation tasks with a lower prioritization value can be shifted in time in order to be able to utilize power generators at a later point in time.

Alternatively or additionally, the power supply parameter is preferably an operating point of at least one power generator for the power grid. A plurality of operating points of a plurality of power generators are preferably considered. An overall operating point of all power generators connected to the electrical network is preferably considered.

A power generator can, in particular, be an internal combustion engine which is drivingly connected to an electrical machine operated as a generator, the electrical machine being connected to the electrical network. Such a combination of internal combustion engine and electric machine is also referred to as a genset. However, such a power generator can also be an electrochemical cell, in particular a fuel cell, a battery, a photovoltaic system, or another suitable power generation or power supply device, in particular in a DC voltage network, but not limited to it.

According to a further development of the invention it is provided that the network stability parameter is selected from a group consisting of a network frequency in the electrical network, a current strength in the electrical network, an electrical voltage in the electrical network, a memory state of at least one with the electrical network connected energy storage, and a target load for the data center. The parameters mentioned here enable the stability of the electrical network to be assessed in a particularly favorable manner. In particular, the network frequency and the electrical voltage are characteristic of the power currently available in the electrical network in relation to the power currently consumed. The grid stability parameter indicates that If the power provided falls in comparison to the power taken off, calculation tasks with a lower prioritization value and / or higher calculation effort are preferably postponed in the data center in order to stabilize the electrical network.

If, on the other hand, the network stability parameter indicates that the provided power tends to exceed the accepted power, additional calculation tasks with a lower prioritization value and / or higher calculation effort - in particular in addition to calculation tasks with a higher prioritization value and / or lower calculation effort - are processed in the data center in order to deal with the electrical Stabilize network. In contrast, calculation tasks with a higher prioritization value and / or lower calculation effort are preferably always carried out - as far as possible - or, if necessary, brought forward.

If the electrical network is an alternating voltage network, the network frequency in particular is characteristic of the power currently available in the electrical network in relation to the power that is currently consumed. If the electrical network is a DC voltage network, in particular at least one of the other network stability parameters mentioned is used.

A storage state of an energy store is understood to mean, in particular, a storage level of the energy store, that is to say a measure of the energy present in the energy store, in particular in relation to a maximum storage capacity of the energy store. Such an energy store can in particular be an electrical store, preferably an electrochemical store, in particular an accumulator or a battery. Such an energy store can, however, also be a water store, in particular a pumped storage power plant, a thermal store, or another suitable storage device. The energy store is connected to the electrical network, possibly with further devices, and is set up to store electrical energy from the electrical network or to deliver it into the electrical network. The energy itself does not have to be stored electrically or electrochemically.

Calculation tasks with a lower prioritization value and / or higher calculation effort are preferably carried out in addition to calculation tasks with a higher prioritization value and / or lower calculation effort if the memory fill level is a problem predetermined first, higher limit value, for example 75% of the storage capacity, has exceeded. In contrast, calculation tasks with a lower prioritization value and / or higher calculation effort are preferably postponed, i.e. only calculation tasks with a higher prioritization value and / or lower calculation effort are processed if the memory level falls below a second, lower limit value, for example 30% of the storage capacity. By intelligently controlling the computing power of the data center, this advantageously enables a reduction in the storage capacity of the energy store used, since a smaller amount of energy has to be kept available, that is to say has to be stored.

A target load for the computer center is understood to mean, in particular, a target value specified by the operator of the electrical network for a power consumption from the electrical network for the computer center. This means that the operator of the electrical network can actively use the data center to regulate network stability. Completely analogous to the above, more calculation tasks can also be calculated with a lower prioritization value and / or higher calculation effort if a higher target load is specified, with calculation tasks with a lower prioritization value and / or higher calculation effort being postponed, or only calculation tasks with a higher prioritization value and / or lower computational effort if a lower target load is specified.

According to a further development of the invention, it is provided that a power prognosis is created for the electrical network, the calculation tasks additionally being processed taking into account the power prognosis. This advantageously allows predictive control of the data center, in particular with a view to the stability of the electrical network. Such a power prognosis can be created in particular on the basis of weather data, in particular a weather forecast, if the electrical network is assigned volatile sources as power generators that provide power depending on the weather, in particular power generators that can be assigned to so-called regenerative energies, for example photovoltaic or wind turbines. However, such a performance forecast can also - additionally or alternatively - be created with a view to a storage state of an energy store of the electrical network. As an alternative or in addition, such a performance forecast can be made with a view to holy days, weekends, School holidays, general calendar events, times of day, behavior of consumers on the electrical network, for example when using means of communication or media, public events such as elections, publication of news or stock market data, and / or much more.

In particular, the chronological order of the processing of the authorization tasks, preferably including a parallel processing of the processing of calculation tasks, is determined as a function of the performance forecast.

According to a further development of the invention, it is provided that a schedule for processing the calculation tasks is set up on the basis of the performance forecast. Thus, the processing of the calculation tasks is advantageously planned for the future on the basis of the performance forecast. Particularly preferably, the schedule for processing the calculation tasks is optimized on the basis of the performance forecast.

According to a further development of the invention, it is provided that outstanding processing tasks, preferably including the at least one assigned parameter, i.e. the assigned prioritization value and / or the assigned calculation effort, are transmitted by the computer center to the electrical network, in particular to an operator of the same. A feedback from the data center to the electrical network then advantageously takes place, which benefits the planning of the stable operation of the electrical network. The operation of slowly controllable power generators, which are connected to the electrical network, is preferably adapted as a function of the outstanding calculation tasks transmitted. In particular, power generators that are to be regulated slowly can be regulated down in good time if it is foreseeable that the data center will draw less power in the future. On the other hand, power generators that are to be controlled slowly can be started up in good time if it is foreseeable that the data center will have a higher power requirement.

The data center preferably has a plurality of computing devices connected to one another in a data network, in particular a plurality of computers or servers. Communication between the individual computing devices can take place via the data network, for example via Wake Up On LAN, or alternatively via a separate one Control line. The computing center advantageously has a higher-level controller, for example a master computing device or a master server, which can switch on or off individual computing devices or groups of computing devices and / or distribute the computation tasks to the various computing devices. Groups of computing devices are preferably switched off by switching a switch or router. Alternatively or additionally, it is possible for individual computing devices or groups of computing devices to be switched off and on by switching switchable power sockets.

According to a further development of the invention it is provided that in a preferred embodiment a cycle of at least one computing device of the data center, preferably a cycle of a plurality of computing devices, preferably a cycle of all computing devices, is influenced as a function of the power provision parameter. This represents a simple as well as elegant and efficient possibility of controlling the processing of calculation tasks and influencing the energy consumption of the computing devices. Behind a clock of a computing device, in particular, a computing clock, in particular a processor clock, is to be understood. A clock signal for the operation of a volatile data memory, in particular a dynamic RAM, that is to say in particular a memory module with random access, is preferably not influenced as a function of the power provision parameter in order to avoid data loss. In this respect, it has been shown that influencing the computing cycle or processor cycle, in particular, leads to a significant variation in energy consumption.

According to a further development of the invention, it is provided that the cycle is selected to be higher or lower as a function of the power provision parameter. In particular, more computing steps can be carried out per time unit if the power supply parameter indicates that more power can be consumed, while fewer computing steps can be carried out per time unit if the power supply parameter indicates that less power should be consumed. Alternatively, it is possible that the clock is switched on or off as a function of the power provision parameter. The energy consumption can thus be influenced in particular in a binary manner by at least one computing device or also a plurality of computing devices, up to all of them Computing facilities of the data center, either calculations performed or no calculations performed.

The clock is preferably switched on or off by a logical rounding, i.e. in particular a logical conjunction, of a clock signal generated by a clock generation with a control signal, in particular a binary control signal, which is generated by the control of the data center as a function of the power provision parameter.

On the one hand, the clock signal and, on the other hand, the control signal as input values in a rounding element, also referred to as an AND gate, the output value of the AND gate being fed to the processor of a computing device as an effective clock signal. Thus, when the control signal is logically high, in particular 1, the clock signal of the clock generation is forwarded to the processor through the AND gate. If, on the other hand, the control signal is logically low, in particular 0, the AND gate blocks the forwarding of the clock signal coming from the clock generation so that the processor is no longer fed a clock signal. The computing device having the processor therefore no longer carries out any calculations.

As already stated, in order to avoid data loss, the clock signal, which is preferably likewise generated by the clock generation, is permanently fed to a volatile data memory, in particular not fed via a corresponding AND gate.

Influencing the clock signal as a function of the power supply parameter has the advantage that the power consumption by the data center can be reduced or increased very quickly, in particular in the low ms range, in particular as a function of a control speed of network parts of the computing devices.

In a further embodiment, the clock can also be influenced by a software command and / or a software message.

According to a further development of the invention, it is provided that cooling of the data center is controlled as a function of the power provision parameter. Thus, the cooling of the data center, which typically has a high demand for electrical power, is also advantageous in the regulation of the network stability and / or the Influencing the efficiency of at least one power generator for the electrical network included. The cooling power is advantageously increased if the power supply parameter indicates that more power can be consumed, while the cooling power is reduced if the power supply parameter indicates that less power should be consumed.

Particularly preferably, a predetermined target temperature range is defined for the cooling of the data center, which is characterized by a lower temperature limit and an upper temperature limit. Depending on the power supply parameters, the cooling is now preferably operated in such a way that the data center - with higher possible power consumption - is operated at the lower temperature limit of the predetermined target temperature range, with lower available power at the upper temperature limit of the predetermined target temperature range. Temperature range is operated. The cooling capacity can thus be increased in times of excess power and reduced in times of insufficient power. In particular, a reduction in the cooling capacity is advantageously possible in that the data center is previously operated at the lower temperature limit of the predetermined target temperature range when there is excess capacity, so that when the cooling capacity is reduced, it only gradually approaches the upper temperature limit.

The object is also achieved by creating a controller, that is to say a control device, which is set up to control a data center operated on an electrical network. The controller is set up to control the processing of a plurality of calculation tasks by the data center as a function of at least one parameter assigned to each calculation task, selected from a prioritization value and a calculation effort, and as a function of a power supply parameter of the electrical network. The controller is set up in particular to carry out the method according to the invention or a preferred embodiment of the method. In connection with the control, the advantages that have already been explained in connection with the method are realized in particular.

The controller preferably has a receiving means for receiving the respective at least one characteristic variable, that is to say the prioritization value and / or the respective one Computational effort. As an alternative or in addition, the controller preferably has an allocation means which is set up to allocate at least one parameter, selected from the prioritization value and the calculation effort, to the calculation tasks.

As an alternative or in addition, the receiving means is preferably set up to receive the service provision parameter.

The controller is preferably set up to control the processing of the plurality of calculation tasks on a plurality of computing devices in the data center, in particular to specify a time sequence and / or parallelization of the calculation tasks on the computing devices. In particular, the controller is preferably set up to specify which calculation task is calculated at which point in time, in particular on which computing device.

In particular, the controller is preferably set up to distribute the calculation tasks to the individual computing devices.

The control is preferably operatively connected to the computing devices via a data network.

The control is preferably also set up to control the computing devices, in particular to switch them on or off individually or in groups, in particular to stop or start a calculation in the computing devices, individually or in groups. It is possible that the controller controls the computing devices via the data network. It is also possible, however, for the controller to be connected to the computing devices via a separate control line, in particular in order to influence a cycle of the computing devices - individually or in groups. The controller is particularly preferably set up to transmit a — preferably binary — control signal to the computing devices via the control line. The controller is preferably set up to generate a - preferably binary - control signal for controlling the at least computing devices, preferably a control signal for each computing device.

The controller preferably has a clock generator, that is to say a clock generator which is set up to generate a clock signal.

The controller also preferably has an AND gate, which is set up to receive the binary control signal and the clock signal and, in particular, to link them to one another in the sense of a logical conjunction To output clock signal - in particular as a result of the logical conjunction - to at least one processor of a computing device. If the binary control signal is now logically high, the clock signal is passed on as an effective clock signal by the rounding element to the processor, so that the processor is clocked and therefore works. If, on the other hand, the binary control signal is logically low, the clock signal is not passed on by the rounding element, so that the effective clock signal disappears. The processor then does not perform any calculations since it does not receive a clock signal. The controller preferably has such a rounding-off element for each computing device.

The object is finally also achieved by creating a data center which has a controller according to the invention or a controller according to one of the exemplary embodiments described above, and / or which is set up for operation on an electrical network according to the method according to the invention or according to a preferred embodiment of the Procedure. In connection with the data center, the advantages that have already been explained in connection with the method are realized in particular.

The computing center preferably has at least one computing device, preferably a plurality of computing devices connected to one another and preferably to the controller, in particular in a data network.

The invention is explained in more detail below with reference to the drawing. Show: FIG. 1 shows a schematic representation of an exemplary embodiment of a data center;

Figure 2 shows a schematic representation of a first embodiment of a

Arrangement of the data center on an electrical network and, at the same time, a first embodiment of a method for operating the data center on the electrical network;

FIG. 3 shows a schematic representation of a second exemplary embodiment of an arrangement of the data center on an electrical network and, at the same time, a second embodiment of the method for operating the data center on the electrical network, and FIG

FIG. 4 shows a schematic detailed illustration of the data center and of the method, in particular according to the first or the second embodiment.

1 shows a schematic representation of a data center 1 which is operated on an electrical network 3 shown in FIG. 2 - in particular as a consumer. The computing center 1 has a plurality of computing devices 5, which can in particular be designed as servers. In addition, the computing center 1 has a controller 7, that is to say a control device, which can be provided in addition to the computing devices 5 or integrated into one of the computing devices 5, in particular a master computing device. The controller 7 is preferably connected to the computing devices 5 via a data network 9. The computing devices 5 are also preferably connected to one another via the data network 9. The controller 7 is set up to control the processing of a plurality of calculation tasks 11 on the computing devices 5, in particular to specify a time sequence and / or parallelization of the calculation tasks 11 on the computing devices 5. In particular, the controller 7 is set up to specify which calculation task 11 is calculated at which point in time, in particular on which computing device 5.

In particular, the controller 7 is preferably set up to distribute the calculation tasks 11 to the individual computing devices 5. The controller 7 is preferably also set up to control the computing devices 5, in particular to switch them on or off individually or in groups, in particular to stop or start a calculation in the computing devices 5 - individually or in groups. It is possible that the controller 7 controls the computing devices 5 via the data network 9. However, it is also possible that the controller 7 is connected to the computing devices 5 via a separate control line 13, in particular in order to influence a cycle of the computing devices 5 - individually or in groups. The controller 7 is particularly preferably set up to transmit a — preferably binary — control signal to the computing devices 5 via the control line 13.

The data center 1, in particular the controller 7, is set up to carry out a method described in more detail below.

Each calculation task 11 is assigned a prioritization value and / or a calculation effort, i.e. at least one parameter, in particular from a client of the data center 1, or a prioritization value and / or a calculation effort is assigned to each calculation task 11, preferably by the controller 7 assigned to at least one parameter.

The controller 7 is set up to specify the processing of the calculation tasks 11 taking into account the respective prioritization value and / or calculation effort, and taking into account a power supply parameter 15 of the electrical network 3. In particular, the controller 7 is set up to determine the chronological order of the processing of the calculation tasks 11, and preferably their parallelization, depending on the respective prioritization value and / or calculation effort, and depending on the service provision parameter 15.

The power supply parameter 15 is preferably a network stability parameter or an efficiency of a power generator for the electrical network 3. FIG. 2 shows a schematic illustration of a first exemplary embodiment of an arrangement of the data center 1 on an electrical network 3 and a first embodiment of a method for operating the data center 1 on the electrical network 3.

Identical and functionally identical elements are provided with the same reference symbols in all figures, so that reference is made to the preceding description in each case.

A plurality of power generators 17 are assigned to the electrical network 3, which in particular can be designed at least partially in different ways, with the electrical network 3 also being assigned to volatile power generators, in particular from the field of regenerative energies. For example, at least one of the power generators 17 can be designed as a wind power plant. At least one other power generator 17 can be designed as a solar or photovoltaic system. It is also possible for at least one of the power generators 17 to be designed as a combination of an internal combustion engine with an electrical machine that is drive-actively connected to the internal combustion engine and operated as a generator, in particular as a so-called genset.

Furthermore, an energy store 19 is preferably assigned to the electrical network 3, which is set up to store energy from the electrical network 3, regardless of the specific physical form of energy storage, and to deliver energy to the electrical network 3. In a preferred embodiment, the energy store 19 is designed as an electrochemical store, in particular as an accumulator or battery.

The data center 1 is connected to the electrical network 3 via an electrical operative connection 21. The electrical operative connection 21 is preferably designed as a cable or line, or as a plurality of cables or lines.

In the first exemplary embodiment shown here and in the first embodiment of the method, the power provision parameter 15 is in particular a network stability parameter. The data center 1 can thus advantageously be used to stabilize the electrical network 3. In general, the network stability parameter is preferably selected from a group consisting of a frequency, a current intensity, an electrical voltage, each preferably measured in or on the electrical operative connection 21, a storage state, in particular storage level, of the energy store 19, and a target load for the data center 1 .

In the embodiment shown here or the embodiment of the method shown, the network stability parameter is preferably a frequency 23 that is measured or picked up at the operational electrical connection 21.

The controller 7 is now set up in particular to postpone the processing of calculation tasks with a lower prioritization value and / or higher calculation effort, depending on a deviation of the frequency 23 from a target frequency for the electrical network 3, and thus to postpone the computing power of the data center 1 and the decrease in power the electrical network 3, or in addition to calculation tasks with a higher prioritization value and / or lower calculation effort also calculation tasks with a lower prioritization value and / or higher calculation effort to calculate the computing power of the data center 1 and thus at the same time its power consumption from the electrical network 3 to increase. In particular, if the frequency 23 falls below the target frequency, the controller 7 reduces the computing power of the data center 1 and thus at the same time the power consumption from the electrical network 3. If the frequency 23 rises above the target frequency, the controller 7 increases the computing power of the Data center 1 and thus its power consumption from the electrical network 3. The data center 1 is thus used in the context of the method as at least a virtual control power source and as a control power sink in order to stabilize the electrical network 3.

When determining the processing of the calculation tasks 11, the controller 7 preferably also takes into account a performance prognosis 25 that is created for the electrical network 3. This power prognosis 25 can be created in particular on the basis of weather data and / or the storage level of the energy store 19. Alternatively or additionally, such a performance forecast can be made with a view to public holidays, weekends, school vacations, general calendar events, times of day, behavior of consumers on the electrical network, for example when using means of communication or media, public Events such as elections, releases of news or stock market data, and / or much more.

The controller 7 particularly preferably sets up a schedule for the processing of the calculation tasks 11 on the basis of the power prognosis 25; it particularly preferably optimizes the schedule on the basis of the power prognosis 25. Thus, the future processing of calculation tasks 11 can advantageously be based on the expected performance of the electrical network 3 be matched.

The controller 7 preferably transmits a feedback 27 regarding outstanding calculation tasks, preferably including their respective at least one parameter, that is, prioritization value and / or calculation effort, to the electrical network 3 or to an operator of the electrical network. This feedback 27 is then preferably used in order to adjust power generators 17 of the electrical network 3, which are to be adjusted slowly, in accordance with the expected future power consumption by the computer center 1.

The data center 1 is also preferably assigned a cooling system 29, which is preferably controlled by the controller 7 also as a function of the power provision parameter 15. The cooling 29 can thus advantageously also be included in the stabilization of the electrical network 3.

3 shows a schematic representation of a second exemplary embodiment of an arrangement of the data center 1 on an electrical network 3 and a second embodiment of the method for operating the data center 1 on the electrical network 3. In this second embodiment or the second embodiment of the method, the power supply parameter is 15 an efficiency of the power generators 17, in particular an overall efficiency of the power generators 17, in particular an overall efficiency of all power generators 17. In this case, the electrical network 3 preferably has a plurality of power generators 17, the efficiency of which depends on the power generated or consumed . All power generators 17 of the electrical network 3 are preferably such power generators, the efficiency of which depends on the power generated or consumed. Such a power generator 17 is for example a genset, but also an electrochemical cell, in particular a fuel cell, or a battery.

The processing of the calculation tasks 11 is now specified by the controller 7 in such a way that the power generators 17 or at least the currently activated power generators 17 are operated at their optimal efficiency or at least close to their optimal efficiency. In particular, the overall efficiency of the power generators 17 is preferably optimized by appropriate planning of the processing of the calculation tasks 11 by the controller 7.

Here, too, feedback 27 is preferably sent to the electrical network 3. In particular, it is possible for the controller 7 to be able to switch individual power generators 17 on or off.

In particular in this exemplary embodiment according to FIG. 3, the electrical network 3 is preferably designed as an island network and assigned to the data center 1 as the sole consumer. In the exemplary embodiment according to FIG. 2, too, the electrical network 3 can be designed as an island network. However, it is also possible for the electrical network 3 to be designed as a supra-regional electrical network, in particular as an interconnected network. In particular, the configuration according to FIG. 3 is suitable for an electrical network 3 designed as a DC voltage network.

4 shows a schematic representation of a detail of the data center 1 and the method for operating the data center 1 on the electrical network 3, this detail correspondingly in the data center 1 according to FIG can be implemented in the second embodiment according to FIG.

Specifically, FIG. 4 shows a particularly preferred embodiment of how the calculation of calculation tasks 11 in one of the computing devices 5 can be started or stopped using the control line 13, in particular using a binary control signal 31 transmitted via the control line 13. For this purpose, the binary control signal 31 is rounded off in a rounding element 33 (AND gate) with a clock signal 37 generated by a clock generator 35, the rounding element 33 outputting an effective clock signal 39 to a processor 41 of the computing device 5. If the binary control signal 31 is now logically high, the clock signal 37 is forwarded as an effective clock signal 39 by the rounding element 33 to the processor 41, so that the processor 41 is clocked and thus works. If, on the other hand, the binary control signal 31 is logically low, the clock signal 37 is not passed on by the rounding element 33, so that the effective clock signal 39 disappears. The processor 41 then does not perform any calculations since it does not receive a clock signal.

In this way, the energy consumption of the computing device 5 can be reduced or increased very efficiently and at the same time quickly, and the processing of calculation tasks in the computing device 5 can be started or stopped. The clock generation 35 generates the clock signal 37 preferably also for a volatile one

Data memory 43, in particular a dynamic RAM. The clock signal 37 is not stopped or prevented for the volatile data memory 43 in order to avoid data loss, in particular by continuing to operate a refresher mechanism. In general, the controller 7 is preferably set up to influence the clock rate of at least one computing device 5 as a function of the power provision parameter 15, in particular to select the clock rate higher or lower as a function of the power provision parameter 15 or - as specifically shown here - to switch it on or off .

Claims

EXPECTATIONS

1. A method for operating a data center (1) on an electrical network (3), wherein

- A plurality of calculation tasks (11) are processed by the computer center (1), wherein

- Each calculation task (11) of the plurality of calculation tasks (11) is assigned a prioritization value and / or a calculation effort, wherein

- The calculation tasks (11) are processed taking into account their respective prioritization value and / or calculation effort, and taking into account a power supply parameter (15) of the electrical network (3).

2. The method according to claim 1, characterized in that the power provision parameter (15) is a network stability parameter.

3. The method according to any one of the preceding claims, characterized in that the power supply parameter (15) is an efficiency and / or an operating point of at least one power generator (17) for the electrical network (3).

4. The method according to any one of the preceding claims, characterized in that the network stability parameter is selected from a group consisting of a frequency, a current strength, an electrical voltage, a storage state of an energy store (19), and a target load for the data center (1) .

5. The method according to any one of the preceding claims, characterized in that a performance prognosis (25) for the electrical network (3) is created, the calculation tasks (11) additionally being processed taking into account the performance prognosis (25).

6. The method according to any one of the preceding claims, characterized in that a schedule for processing the calculation tasks (11) based on the performance forecast (25) is set up, preferably optimized.

7. The method according to any one of the preceding claims, characterized in that outstanding calculation tasks (11), preferably with the respectively assigned prioritization value and / or calculation effort, are transmitted through the data center (1) to the electrical network (3).

8. The method according to any one of the preceding claims, characterized in that a cycle of at least one computing device (5) of the data center (1) is influenced as a function of the power provision parameter (15).

9. The method according to any one of the preceding claims, characterized in that the clock rate is selected to be higher or lower, or switched on or off, as a function of the power provision parameter (15).

10. The method according to any one of the preceding claims, characterized in that a cooling (29) of the data center (1) is controlled as a function of the power provision parameter (15).

11. Controller (7), set up to control a data center (1) operated on an electrical network (3), the controller being set up to allow the data center (7) to process a plurality of calculation tasks (11) as a function of one each computation task (11) respectively assigned prioritization value and / or from a computation effort each assigned to each calculation task (11), and to be controlled as a function of a power supply parameter (15) of the electrical network (3).

12. Data center (1), with a controller according to claim 11.